Background: Pancreatic cancer is the fourth leading cause of cancer death in the United States. Surgery offers the only chance at cure however, less than 20% are resectable at presentation. A common reason for being classified as unresectable is loco-regional advanced disease. Several phase-I studies of regional administration of chemotherapy have been proven safe. The main advantage of pancreatic cancer targeted arterial infusion of Gemcitabine is achievement of higher local bio-available active drug levels at the tumor bed. The RECLAP trial will offer super selective prolonged administration of Gemcitabine via a subcutaneous port for patients with unresectable locally-advanced pancreatic cancer. Objectives: Primary Objective: To evaluate feasibility and toxicity. To establish the MTD Secondary Objectives: To evaluate response rate using RECIST, PET, Functional MRI and CT perfusion criteria To evaluate the conversion rate from unresectable to potentially resectable pancreatic cancer. To determine progression free and overall survival. Eligibility: Locally advanced (unresectable) pancreatic cancer. 18 years old or greater with an ECOG 0-2 Laboratory and physical examination parameters within acceptable limits by standard of practice guidelines prior to surgery Design: 12 patients with unresectable locally-advanced pancreatic cancer will receive super-selective continuous arterial infusion of Gemcitabine over 12 hours. Initial dose will be 300mg/m2 (20% of the normal I.V. dose);treatment will be given every 2 weeks (one cycle = 4 weeks). Every cycle the dose will be increased (300, 600, 1000 and 1200m2) unless grade-2 pancreatic specific toxicity ensues. 12 patients will be divided into 3 groups. Group-1 will start on 300mg/m2, group-2 on 600mg/m2 and group-3 on 1000mg/m2, unless previous DLT. Patients will be evaluated every 2 cycles. Upon progression patients will be taken off study but continued to be followed. If no PD patients will continue up to six cycles.